Towards clinical application of liver, vessel, and tumor segmentation using partially labeled data

Eirik Agnalt Østmo, Keyur Radiya, Kristoffer Knutsen Wickstrøm, Michael Kampffmeyer, Karl Øyvind Mikalsen, Robert Jenssen
Proceedings of the 7th Northern Lights Deep Learning Conference (NLDL), PMLR 307:415-427, 2026.

Abstract

Accurate delineation of liver parenchyma, intrahepatic vessels, and tumors (LVT) may aid earlier tumor detection, consistent response assessment, and surgical planning for patients with liver cancer. Deep learning (DL) may enable such automated delineation, but available CT datasets are inconsistent and partially labeled, making them unsuited for end-to-end training. We investigate a single-head, 3D segmentation framework that learns from partially labeled data by: (i) loss masking per class or voxel to ignore missing annotations, (ii) using multi-hot targets and the anatomical hierarchy inherent to liver, vessels, and tumors, to handle overlapping structures without class competition. In controlled ablations that simulate partial-label training, this multi-label masked strategy reliably outperforms masked multi-class baselines, avoids precision collapse, and improves tumor overlap and lesion detection sensitivity. Scaling training to multiple partially labeled datasets, the model surpasses full-resolution nnU-Net on an external clinical cohort, with higher tumor and vessel segmentation performance. We conduct a retrospective feasibility analysis on clinical data to illustrate the clinical potential of the LVT application. We find that LVT models may facilitate earlier detection of metastasis, longitudinal size tracking aligned with radiologist measurements, 3D tumor–vessel visualization for surgical planning, and stable inter-phase liver volumetry ($\approx$ 5% deviation). These results show that multi-label masked learning enables robust, clinically relevant LVT segmentation from partially labeled datasets.

Cite this Paper

BibTeX
@InProceedings{pmlr-v307-ostmo26a, title = {Towards clinical application of liver, vessel, and tumor segmentation using partially labeled data}, author = {{\O}stmo, Eirik Agnalt and Radiya, Keyur and Wickstr{\o}m, Kristoffer Knutsen and Kampffmeyer, Michael and Mikalsen, Karl {\O}yvind and Jenssen, Robert}, booktitle = {Proceedings of the 7th Northern Lights Deep Learning Conference (NLDL)}, pages = {415--427}, year = {2026}, editor = {Kim, Hyeongji and Ramírez Rivera, Adín and Ricaud, Benjamin}, volume = {307}, series = {Proceedings of Machine Learning Research}, month = {06--08 Jan}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v307/main/assets/ostmo26a/ostmo26a.pdf}, url = {https://proceedings.mlr.press/v307/ostmo26a.html}, abstract = {Accurate delineation of liver parenchyma, intrahepatic vessels, and tumors (LVT) may aid earlier tumor detection, consistent response assessment, and surgical planning for patients with liver cancer. Deep learning (DL) may enable such automated delineation, but available CT datasets are inconsistent and partially labeled, making them unsuited for end-to-end training. We investigate a single-head, 3D segmentation framework that learns from partially labeled data by: (i) loss masking per class or voxel to ignore missing annotations, (ii) using multi-hot targets and the anatomical hierarchy inherent to liver, vessels, and tumors, to handle overlapping structures without class competition. In controlled ablations that simulate partial-label training, this multi-label masked strategy reliably outperforms masked multi-class baselines, avoids precision collapse, and improves tumor overlap and lesion detection sensitivity. Scaling training to multiple partially labeled datasets, the model surpasses full-resolution nnU-Net on an external clinical cohort, with higher tumor and vessel segmentation performance. We conduct a retrospective feasibility analysis on clinical data to illustrate the clinical potential of the LVT application. We find that LVT models may facilitate earlier detection of metastasis, longitudinal size tracking aligned with radiologist measurements, 3D tumor–vessel visualization for surgical planning, and stable inter-phase liver volumetry ($\approx$ 5% deviation). These results show that multi-label masked learning enables robust, clinically relevant LVT segmentation from partially labeled datasets.} }
Endnote
%0 Conference Paper %T Towards clinical application of liver, vessel, and tumor segmentation using partially labeled data %A Eirik Agnalt Østmo %A Keyur Radiya %A Kristoffer Knutsen Wickstrøm %A Michael Kampffmeyer %A Karl Øyvind Mikalsen %A Robert Jenssen %B Proceedings of the 7th Northern Lights Deep Learning Conference (NLDL) %C Proceedings of Machine Learning Research %D 2026 %E Hyeongji Kim %E Adín Ramírez Rivera %E Benjamin Ricaud %F pmlr-v307-ostmo26a %I PMLR %P 415--427 %U https://proceedings.mlr.press/v307/ostmo26a.html %V 307 %X Accurate delineation of liver parenchyma, intrahepatic vessels, and tumors (LVT) may aid earlier tumor detection, consistent response assessment, and surgical planning for patients with liver cancer. Deep learning (DL) may enable such automated delineation, but available CT datasets are inconsistent and partially labeled, making them unsuited for end-to-end training. We investigate a single-head, 3D segmentation framework that learns from partially labeled data by: (i) loss masking per class or voxel to ignore missing annotations, (ii) using multi-hot targets and the anatomical hierarchy inherent to liver, vessels, and tumors, to handle overlapping structures without class competition. In controlled ablations that simulate partial-label training, this multi-label masked strategy reliably outperforms masked multi-class baselines, avoids precision collapse, and improves tumor overlap and lesion detection sensitivity. Scaling training to multiple partially labeled datasets, the model surpasses full-resolution nnU-Net on an external clinical cohort, with higher tumor and vessel segmentation performance. We conduct a retrospective feasibility analysis on clinical data to illustrate the clinical potential of the LVT application. We find that LVT models may facilitate earlier detection of metastasis, longitudinal size tracking aligned with radiologist measurements, 3D tumor–vessel visualization for surgical planning, and stable inter-phase liver volumetry ($\approx$ 5% deviation). These results show that multi-label masked learning enables robust, clinically relevant LVT segmentation from partially labeled datasets.
APA
Østmo, E.A., Radiya, K., Wickstrøm, K.K., Kampffmeyer, M., Mikalsen, K.Ø. & Jenssen, R.. (2026). Towards clinical application of liver, vessel, and tumor segmentation using partially labeled data. Proceedings of the 7th Northern Lights Deep Learning Conference (NLDL), in Proceedings of Machine Learning Research 307:415-427 Available from https://proceedings.mlr.press/v307/ostmo26a.html.

Related Material